Wolfspeed, Inc. (WOLF) Earnings Call Transcript

Welcome. My name is Jed Dorsheimer, Global Head of Sustainability Research at Canaccord Genuity. And for those who have been with me this morning, welcome back. We continue on the journey as we've been kind of taking some of the higher-level abstract concepts. And now the goal is to apply that to 3 different domains. One, of silicon carbide and the use of silicon nitride, both in traction inverters as well as some other applications. So moving up the higher electron mobility, if you will. And then after this, we will have Fusion and then the other wide-band-gap semiconductor, gallium nitride. And so I'm just delighted to have with me 2 leading experts in the area. So I have John Palmour -- Dr. John Palmour, Co-Founder of Cree. He and I have known each other for quite a while. He is leading the Wolfspeed business and focusing on and pioneering high-electron mobility semiconductor technology and power as well as RF and commercializing that from LEDs over to power semis. And it's been a -- I can say firsthand, it's been an interesting journey as I knew you back when in the early LED and now today and seeing what you guys are doing, and it's just been a pleasure and an honor to watch. Bob Daigle, who I've met more recently, and we recently picked up coverage of Rogers Corporation. And he is Chief Technology Officer over at Rogers. And Rogers is fairly broad-based and diversified in terms of what they do, but a particular area of interest that Bob and I tend to talk about more often than not is that of the silicon nitride technology, which is a -- which is used as a heat sink to extract the heat generated from John's products in applications like traction inverters. So I guess thank you is where to start for both of you being with us. I'm not sure if either of you have caught some of the previous presentations, but we've talked a lot about concepts of biophysical economics and freedom of flow. Adrian Bejan from down in Duke, your neck of the woods, John, was with us and talking about construct the law. And I think all of this is kind of leading and I'm sure I'm going to get from investors, so what? So what are these concepts really matter? And that's where I'd like to -- well, before I maybe jump in the deep end there, we'll take a couple of minutes to maybe describe your backgrounds and the companies as you see fit. So John, I'll turn it to you if you'd like to maybe add to anything that I missed there and describe Wolfspeed a bit further, and then I'll do the same with Bob.

Sure. So Cree is about 33-, 34-years-old, we formed in 1987. We were a spin-off from North Carolina State University, where we were working on silicon carbide. And in the ensuing 34 years, we've been dedicated to commercializing wide-band-gap semiconductors, primarily silicon carbide or gallium nitride on a silicon carbide substrate. As Jed had mentioned, we were very, very active in LEDs, blue and white LEDs, and then LED lighting more than 20 years ago. And more recently, we've really been emphasizing silicon carbide for power devices in both industrial markets and also for automotive and gallium nitride on silicon carbide RF devices. But the common theme through all of these product pushes that we've had from LEDs to power to RF, the common theme is energy savings, higher efficiency. We always like to say that the cheapest energy -- the cheapest form of alternative energy is the energy you never burn. So pretty much our consistent value proposition has been saving the end customer energy. So silicon carbide is now getting a lot of interest in the automotive market, which Jed mentioned, and we're putting a lot of efforts towards. And we're the world's leading supplier of silicon carbide material by a good bit and looking forward to explosive growth in the future.

And JP, what's the most energy-efficient LED?

That is a question I wish I knew. I actually don't know.

The one that's off. So Bob -- sorry...

100% climate efficiency there.

So Bob, I want to turn it to you, just same, take a couple of minutes and just frame out how you see Rogers, your role at Rogers and the technology that you're developing?

Sure. Yes. So again, Bob Daigle at Rogers, Chief Technology Officer. And like Jed mentioned earlier, we're a very diverse specialty materials company. But an area that we've really focused on and invested in really the past 6, 7 years has been this whole area of advanced mobility. A lot of our early involvement has been in RF circuit materials for advanced driver assistance systems. So basically, your blind spot detection, adaptive cruise control, where we've developed some products that are used broadly for those radar sensors. But I'd say, the past 4, 5 years, a lot of investment on our side really around this whole area of EV/HEV in a number of areas. We're heavily involved with pad materials, we call them compression pads that are used in these pouch cell batteries. We do a lot in the area of battery interconnects. And the third area of focus, which will be a focus really for today's discussion, is around power semiconductor packaging, especially as it relates to high-power applications and the inverters and the work we've done really to try to really support people like John in terms of their road maps and delivering these wide-band-gap semiconductors that provide a lot of advantages in the -- for electric vehicles, but being able to handle the heat, the thermal requirements and also the reliability requirements in an automotive application, which are much more demanding than what would have been found in a traditional industrial motor drive. So looking forward to today's discussion, I think we're both -- we're playing in a very exciting area, and it's great to be part of the ecosystem that hopefully will -- or is moving very quickly to enabling cleaner, more sustainable environment.

Well, thank you, and thank you both for being here. So the title is Renaissance in the Semiconductor Sector, and why Moore's Law is the Most Irrelevant Discussion Point for Investors. I find that my 20-plus year career in -- both on Wall Street and governance as well as in industry, more often than not, I get asked it, it's the common nomenclature that, well, where are you on Moore's Law? And what's your cost down? And while I think that observation, one, it's not a law, but an observation. And two, it was supposed to only last 10 years and has lasted 40-plus years and been slowing since it started. But provided an excellent understanding for what I talked about earlier, which is most people, including investors, have a hard time on log scale growth. And that we think, and I think most humans are wired in a linear fashion.

I'd love to ask the question, if you could fold a piece of paper 100x, how thick with the ream of paper be, and a few people will say, "Oh, it's going to be greater than the distance of the known universe," but it is. And so you can compound things really quickly, and that's where the doubling of transistors every 18 to 24 months was really important from a cost down. But what was often overlooked in that is the understanding of the Baliga's figure of merit in what electron mobility and doing more with less. And I think the handset guys really got this with RF in terms of gallium arsenide and what you could do. So I guess, one, just open question, thoughts on that? I'll turn to JP first and then Bob second, and then I want to come back to probably a more difficult question.

Yes. So Kind of 2 thoughts on that. One is more Moore's Law does not really apply to power semiconductors. And the reason is you can't make a device infinitesimally small for handling power. Why is that? Well, a lot of it depends on what Bob is going to talk about because that means you got to get rid of more power. So the big advantage for silicon carbide is we can get a lot more amps out of a smaller chip that has lower capacitances, et cetera. So we have lower ohm resistance. But as you shrink that chip, the thermal load is getting put into a smaller and smaller area. So how do you get rid of that? And so they're -- you're up against a wall on thermal limits. And CMOS faced the same thing. It's not like there wasn't some thermal limit. But when you're talking about power devices or RF devices, it's a pretty extreme limit. So Baliga's figure of merit is a really good guide for figuring out what materials have potential. The thing you have to remember with Baliga's figure of merit is it's a theoretical calculation. What it says is if there were no other parasitic resistance and you were completely dominated by the resistance of the drain region, this is the curve of ohm resistance versus voltage that you would have. What we have to deal with in the real world is what are the other parasitic resistances and how do we get rid of those because the whole build is to get closer and closer to Baliga's figure of merit. And we put a lot of effort on reducing the other sources of resistance. And as we make those improvements, it puts more pressure on the packaging to get rid of that thing. So they're both very valid and interesting laws, but they don't necessarily reflect the challenges that we have to deal with in the real world.

Bob, your thoughts?

Yes. So I think -- and I think, on the bigger discussion, which you and I had before, Jed, is this whole context of it's really around system cost. And it's around system performance, efficiency and cost, and to the extent that where Moore's Law basically had a number of scaling factors, which were very beneficial. I think the curves are different, but you have similar opportunities as we address the various challenges in the power semiconductor area. Now one of them was obviously the wafer size got larger, and as Cree's been very local -- I mean, Cree's investing heavily, right? They're putting in larger scale wafer capacity. The other aspect that you're seeing is the ability to operate. As these multigenerational devices have been introduced, you're seeing that you can basically drive them harder. So the same current that you would have gotten per square millimeter years ago, you're doing much better now. I mean I think -- and there's clearly some road maps around driving the energy density challenges. And then to John's point, then, okay, I'm driving this harder, but now I have a lot of heat per unit area that I had to remove. And that's where our side of the tent here in terms of the packaging technology has evolved, where the typical silicon device was running on a substrate that had about less than 1/4 the thermal conductivity we're providing now. So those silicon nitride substrates versus the alumina substrates that were used with all the silicon devices has about 4, 4.5x thermal conductivity. And that's the first generation. There are going to be those 90 watt per meter K substrates are going to evolve to 120, 130-watt per meter degree. So I think you're seeing the different parts of the challenges that are -- that really are important to really drive cost down, where you're getting bigger formats, you're getting more devices per unit area that they're doing the same job. You're removing the heat more efficiently because you've got better thermal interface products. I think you're seeing things come together so that there's a reason to -- and it's not Moore's Law. So maybe we'll have to invent the law. But there will be -- it's very reasonable to believe that you're going to see a solid evolution of costs that are very favorable for the industry. And just one last point is, which is interesting today because the economics already work. So if I'm a carmaker and I'm willing to do system-level engineering, not subcomponent optimization, I can get a lower cost system because I can put a smaller battery into the same range. I can reduce the complexity of my thermal -- my cooling system on the inverter. I can make a smaller inverter. So what's exciting, I think, is you already have economics that work in the industry and very nice road maps for continuing to drive down cost and improve performance.

Yes. And I think one of the important points, John, you could probably provide some context here. We thought a lot of the same things with respect to LED. And the LED industry have this physics phenomenon, and I'm referring to droop, in terms of -- which really took us by surprise because we thought, ultimately, you would -- maybe you wouldn't have a 2-inch LED, but you would optimize to a larger LED. And what happened was you hit a -- you had an asymptotic curve in terms of the lumen output with radiative recombination that allowed okay to be good enough. And one of the important points with power semiconductors is you actually don't have that in terms of electron mobility. And I feel like that's really not well understood by either industry participants or, quite frankly, investors here.

Yes. It's a good point because, yes, the phenomenon you're talking about was the Auger recombination in LEDs as -- which meant that -- I mean, you had a fairly linear output of light versus current, but then it would start to roll off. And so you couldn't just make a big chip, pump a lot of current through it and get a lot more light. And what people started doing was putting relatively cheap, small LEDs in arrays. For power -- so number one, in power devices, we do not have that limitation at all. There are some practical limitations on die size, mostly around yield, thermal cycling, et cetera. But there's no phenomenon like Auger recombination and things. So we're not limited in that respect. But the other phenomenon is you can't just make a bunch of little chips and parallel them because of the inductances of the wires. I mean, we're operating at frequencies where these wire inductances have a major impact on the performance of the power module. So not to mention, if you parallel a bunch of small chips and you put a bunch of wires, wires are the #1 source of failure generally in a power module. So it's pretty impractical actually to parallel a bunch of small devices. So there's a sweet spot in there in terms of device size where you want it big enough to where you don't have 1 million chips in parallel but you don't want it so big that the cost gets too high because of typical yield considerations. And that's the area we operate in. There are some defect in densities in silicon carbide that still limit you. So we don't make chips quite as big as IGBTs, but those -- I don't see us ever getting over probably a 1 centimeter squared chip just due to typical practical yield issues, even in the fab discounting material.

Well, and just for size, I don't know if you want to say it, but I will. In terms of the size comparison to an IGBT, it's roughly about 4 to 5x smaller in terms of getting the same apples-to-apples output in terms of ohmic and voltage and characteristics and specifications. So if I'm in a 200-millimeter wafer, for example, I might be able to get 250 or so IGBTs, and I'm going to get about 1,000 silicon carbide MOSFETs on that same 200 millimeters, that's the apples-to-apples for investors that are looking at the economics.

It's absolutely true. And that's the thing a lot of people don't appreciate when it comes to looking at GaN, both GaN on silicon and -- because it's a lateral device, the chips are actually much bigger. So we get a lot more amps per wafer than you do with GaN on silicon. We get a lot more amps per wafer than you do with IGBTs and silicon. And so that helps offset the higher material cost, et cetera, that we have. And so that's part of why we're able to be competitive in this market. And the other thing, and I think we'll probably get in more to this is people used to very much look at component-to-component pricing. And that was a hard thing to get companies to look at system-level costs, just as Bob said. And you have to get the typical buyer from a large OEM to get past that and look at what does it do for my system overall? Because despite the fact that our device might be more expensive than IGBTs, the system-level cost is significantly lower. And just getting that dialed or that process going took a while, but now I think most -- pretty much everybody understands that.

I think we have a lot to thank Tesla for really pioneering and demonstrating very clear that -- clearly to the industry that the savings -- because that inverter, if it were silicon-based versus silicon carbide chips, you can probably look at maybe reducing system losses by, I've heard as much as 2/3 by going to a wide-band-gap semiconductor fast switching. So there's a lot of efficiency improvement available, but you only get it if you're willing to look at the overall system. Otherwise, if you're trying to piece things together in a more traditional automotive model, you spend the money, but you don't get the benefit. And I think Tesla basically educated the industry about the -- how critical it is. And you see it now. I think to John's point, you're seeing it's a much different discussion today than it would have been a couple of years ago at the OEM level, and there's a lot of activity right now and really pulling together the cross-functional, cross-organizational teams that they need to be able to take advantage of this technology.

Yes. I mean, no doubt that Tesla certainly moved the goalpost. And I would say -- I'm not sure I'd use the word educate, but embarrassed other OEMs in the marketplace. And I think we should -- we all owe a great deal of gratitude for showing what can be done. But it's also -- I think what's amazing is it's not as if that's the optimized approach either. I mean where we're at in terms of design is, to me, sort of like that first Pentium back in the late 1980s, like there's an -- we don't know what we don't know at this point in terms of opportunity. I don't think anybody in the late '80s envisioned what a smartphone -- or a real vision of what a smartphone would be able to do in terms of processing power. And so if we think in terms of the context of electrification and what we'll be able to do in terms of efficiency, it really seems strikingly similar in terms of that context. And I think Intel is probably the apropos comparison there.

So I would say that, yes, we are definitely at the beginning, at the very early stages. The devices that Tesla designed in were fairly early generation to us, and other silicon carbide device providers are all working on the next-generation devices, which are generally going back to the Baliga figure of merit, generally trying to reduce specific ohm resistance, which reduces cost, gives you a smaller chip. But we're also working a lot on ways to extract more value out of the silicon carbide that you have through device design and/or packaging technology. If you were to make typical power modules, the established power module designs that silicon has, you're only getting maybe half the value of silicon carbide. If you can improve the inductance of that module through layout and optimization and they can have a higher thermal conductivity substrate, you can effectively get about double the value out of that power module. So there's work being done all the way from the device design up to the finished module design to optimize that. And yes, I think we're really at the beginning of that journey.

Yes. I think it's -- it probably is incumbent on us to -- in terms of where we're at in that design, not only are we at the beginning, but we haven't even touched on the optionality. I mean if I design with 800 or 900 volts in terms of that system architecture, one has to ask as an engineer, what are the other savings? What is my exposure to copper? What does that do in terms of induction versus asynchronous versus permanent magnet? Oh, well, neodymium is going to be going up considerably as we all think about moving off of hydrocarbons to motors. What is that optionality going to provide and really that -- we're still very myopically looking in the comparison between, oh, here's what an IGBT does, here's what the cost road map looks like on that, here's what a MOSFET does, apples-to-apples, where is the crossover point and where does that scale, which alludes all of those other benefits, which if I -- if we look and use that comparison to traditional silicon ICs in processors, that's where you really had that hyperdrive. It wasn't just about it was what could you do with those extra transistors in a similar ways, what can you do with the extra electrons in this case?

Fine. I would also say that translates all the way up to the design of the vehicle itself. So we're making things much more compact, much more efficient. That frees up space. So for the OEM designing the vehicle, okay, what can I do with that extra space? And that would add value to the car.

Yes. And I think if you marry that along with what's going on in the battery area, and this whole idea of range anxiety, I think, is not far field from being addressed. I think the combination of you talk high voltage, the ability to fast charge these vehicles, the combination of not just the battery technology, but the fact that the inverters are much more efficient means I'm getting that incremental range. So the idea of having these 500-mile vehicles that you can recharge in 10 minutes is not that far away today. And I think -- and the economics are coming down quickly enough where that argument, I think, is also going to be relatively short lived as the different parts of the technology start to -- they're all advancing, but more importantly, I think their -- that system-level engineering, design, integration, I mean, even on the battery space right now, you're seeing that the batteries themselves are going to be the structure of the vehicles, so weight savings. Now 800, 900 volts means I don't have a lot of heavy conductors. It might be [ 400-volt ], they're a lot less heavy than they used to be. So I think the industry has come a long way in 2, 3 years. But I do think, to your point, is this the Pentium processor on the semiconductor, but also quite frankly on the EV, a broader discussion of EV platforms. I think that's true. I think we're very early in the evolution of these vehicle technologies.

Yes. And just to build on that further, basically 2, 3 years ago, almost everybody talked about a 400-volt bus, and not many people were talking about 800-volt. But to Bob's point, it shifted a lot, and a big part of that shift not just the copper savings, but the ability to fast charge. And again, the whole logic and thought process around fast charging, I think, is pretty early. Could you go higher? And there comes a point where you can only dump so much energy into a battery before it explodes. We obviously don't want to go there. But there's -- it is very early in the game, I think.

Yes. I mean there was a period of time with plentiful oil where 5 miles per gallon was highlighted as breakthrough efficiency on a carburated engine. And today, table stakes are sort of 35 miles per gallon, right? And so in a similar way, when more often in the beginning of a trend, you're not going to optimize it for what's down the road.

And you think about where we're at today, right? So in a well -- and by today's standards, a well-engineered vehicle, the efficiency -- so if I -- if you do an apples-to-apples energy for your Tesla versus an equivalent size internal combustion vehicle, you're already 4x more efficient in moving people around. So we're not a long -- I mean that's the starting point. We're -- the technology has moved along, but there's a lot more room to go and we're 4x. So it's very impactful, I think, on efforts to deal with climate change and CO2 emissions.

So I want to pause for one second for the audience members, I see there's a lot of people on. If you do have a question, you can put that in, and I'll take a look and try and ask that -- present that to Bob and John. I do want to throw a bit of a curveball at you, John, and I'll start with you first. I'd like you to take the -- I'd like you to argue why silicon carbide won't happen? Take the steel man perspective of what -- and I know that it's antithetical to what you actually believe, but I do -- I'd like you to kind of put that out there so we could unpack that a little bit. I'm going to ask you, Bob, the same question, I guess, for silicon nitride, if you will?

Yes. So I mean, this is the battle I've fought for a lot of years. And I think it would still be the same. There were 2 main reasons why you wouldn't want to go with silicon carbide. One was cost. And again, that thought process -- and we still deal with it in industrial markets where people tend to look at component and component costs. And the other was reliability. It was a new semiconductor technology. There was a lot of fear about could it be reliable enough. That has, by and large, gone away. And when I say reliability, the fundamental reliability of the technology. I think that issue has been retired by now. And now it's very much more a, what I'll call, a quality question, which is a vendor by vendor issue. What's your fail rate? What are your modes and failure, et cetera, et cetera. That's just typical quality issues that are -- every power semiconductor company has to answer for automotive. But I think the reliability question has been retired. I mean we've had trillions of device hours in the field. It's a very reliable technology. So cost is the big one. And I think, again, that has what -- has been what's changed. Again, a lot of credit to Tesla. In the industrial markets, and I'll share this, we could be in motor drives and save a lot of energy. The way that it happens, though, is if you're building a facility, you're given a budget. And you've got 2 choices, and oh, that one motor drive may be more expensive. Most everybody goes for the cheaper one because paying the power bill is someone else's problem. And so while the automotive business has gotten that message of total system cost at high levels, very high levels, frankly, that doesn't happen in a lot of the industrial markets because of the way things are structured. A guy who's building a project is going to try to meet his budget, and he's not going to worry about the power bill, even if this more expensive motor drive could pay for itself in a year or 2 years or whatever. So that is still more of an issue in industrial markets, but cost is generally the main thing we though of against as a steel man.

Bob, turning to you?

Yes. So I was just thinking a little bit about -- so if John could wave a magic wand and make silicon carbide devices that don't generate any heat, so 100% efficiency, I think -- but the reality is what we do is we remove heat, right? We do it in a reliable way. We have -- and then it gets back to something...

By the way, I want to...

Yes, go ahead, John.